gh-135241: Make unpickling of booleans in protocol 0 more strict (GH-135242)

The Python pickle module looks for "00" and "01" but _pickle only looked
for 2 characters that parsed to 0 or 1, meaning some payloads like "+0" or
" 0" would lead to different results in different implementations.
This commit is contained in:
Justin Applegate 2025-08-14 13:22:37 -06:00 committed by GitHub
parent 0cbbfc4621
commit 781294019d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 14 additions and 1 deletions

View file

@ -1012,6 +1012,16 @@ def test_constants(self):
self.assertIs(self.loads(b'I01\n.'), True)
self.assertIs(self.loads(b'I00\n.'), False)
def test_issue135241(self):
# C implementation should check for hardcoded values 00 and 01
# when getting booleans from the INT opcode. Doing a str comparison
# to bypass truthy/falsy comparisons. These payloads should return
# 0, not False.
out1 = self.loads(b'I+0\n.')
self.assertEqual(str(out1), '0')
out2 = self.loads(b'I 0\n.')
self.assertEqual(str(out2), '0')
def test_zero_padded_integers(self):
self.assertEqual(self.loads(b'I010\n.'), 10)
self.assertEqual(self.loads(b'I-010\n.'), -10)

View file

@ -0,0 +1,3 @@
The :code:`INT` opcode of the C accelerator :mod:`!_pickle` module was updated
to look only for "00" and "01" to push booleans onto the stack, aligning with
the Python :mod:`pickle` module.

View file

@ -5255,7 +5255,7 @@ load_int(PickleState *state, UnpicklerObject *self)
}
}
else {
if (len == 3 && (x == 0 || x == 1)) {
if (len == 3 && s[0] == '0' && (s[1] == '0' || s[1] == '1')) {
if ((value = PyBool_FromLong(x)) == NULL)
return -1;
}